Semantic Query-based Generation of Customized 3D Scenes Krzysztof Walczak, Jakub Flotyński...
38
Semantic Query-based Generation of Customized 3D Scenes Krzysztof Walczak, Jakub Flotyński Department of Information Technology Poznań University of Economics, Poland [walczak, flotynski]@kti.ue.poznan.pl Web3D 2015 – The 20th International Conference on 3D Web Technology June 18-21 – Heraklion, Crete, Greece
Semantic Query-based Generation of Customized 3D Scenes Krzysztof Walczak, Jakub Flotyński Department of Information Technology Poznań University of Economics,
Semantic Query-based Generation of Customized 3D Scenes
Krzysztof Walczak, Jakub Flotyski Department of Information
Technology Pozna University of Economics, Poland [walczak,
flotynski]@kti.ue.poznan.pl Web3D 2015 The 20th International
Conference on 3D Web Technology June 18-21 Heraklion, Crete,
Greece
Slide 2
Presentation Outline Motivations for semantic modelling of 3D
content The SEMIC approach Query-based content customization
Examples Implementation Conclusions and future works 2
Slide 3
Interactive 3D Web Content Opportunities for widespread use of
3D content Increasing performance and versatility of hardware Cheap
3D accelerators and displays Rapid growth in the network bandwidth
Numerous domains Cultural heritage, education, medicine, tourism,
e-commerce, More complex than other types of media Geometry,
structure, space, appearance, animation, behaviour 3
Slide 4
3D Content Creation Challenge Creation of interactive 3D
content is difficult High complexity of content Requires proper
tools and experience Multitude of content presentation platforms
Expensive and time-consuming Applicability of ready-to-use 3D
content is limited Personalization visualization, teaching,
cultural heritage, medicine Adaptation different presentation
platforms and contexts Content creation and customization performed
by domain experts or end users Efficient methods of creation and
customization of 3D content are required! 4
Slide 5
Creation and Customization of 3D Content Solutions simplifying
3D content creation/customization Specialized tools and
applications High-level languages and frameworks Templates,
parameterization, components Limited possibilities of facilitating
content creation/customization 5
Slide 6
Problem statement Lack of solutions for flexible on-demand
content creation/customization including explicit and implicit
management of different aspects of 3D content at different levels
of abstraction 6
Slide 7
Semantic Modeling of 3D Content Extension and generalization of
the previous approaches Declarative and knowledge-based content
creation at arbitrarily chosen (conceptual) level of abstraction
Content representation independent of particular presentation
platforms Simplified content management (indexing, searching and
analysing of content components) 7
Slide 8
The SEMIC Approach 8 Creation of a generalized content
representation Query-based content generation Building a final
content representation Content presentation Semantic 3D meta-scene
Customized content representation Final content representation 3D
Content Representation Language StepsContent modelsContent
representations Semantic Content Model Semantic Content
Customization Patterns
Slide 9
The SEMIC Approach 9 Creation of a generalized content
representation Query-based content generation Building a final
content representation Content presentation Semantic 3D meta-scene
Customized content representation Final content representation 3D
Content Representation Language StepsContent modelsContent
representations Semantic Content Customization Patterns Semantic
Content Model
Slide 10
10 Content creation Conceptual Content Representation
Representation Mapping Concrete Content Representation Semantic
Content Representation Domain-specific ontology Multi-layered
Semantic Content Model (ML-SCM) Semantic Mapping Model (SMM)
Semantic Content Model
Slide 11
Ontology Mapping Semantic Mapping Model Enables presentation of
domain-specific concepts by mapping them to concrete elements
(ML-SCM) Includes generic concepts for wrapping domain-specific
ontologies Presentable objects Properties Descriptors Relations 11
Domain-specific Ontology Multi-layered Content Model (ML-SCM)
Semantic Mapping Model (SMM)
Slide 12
Ontology Mapping Presentable Objects Represent primary
(independent) entities of a scene at a conceptual level of
abstraction Artefacts in a virtual museum exhibition Avatars in a
game Cars in a car show Super-classes of domain-specific classes
with independent representations Reflected by geometrical and
structural 3D content -specific classes 12
granarystatuettestandcart
Slide 13
Ontology Mapping Properties Represent conceptual features of
presentable objects, e.g., material, colour, shininess, etc.
Super-properties of domain-specific properties Reflected by 3D
content-specific properties describing Geometry Structure Space
Appearance Animation 13 granary.madeOf=wood stand.colour=beige
statuette.texture= pict.png
Slide 14
Ontology Mapping Descriptors Containers collecting multiple
properties Descriptive classes containers for properties with
values fixed for a whole class of presentable objects Descriptive
individuals containers for properties with values variable within a
group of objects 14 RotatingObject PlasticObject
TexturedObject
Slide 15
Ontology Mapping Relations Represent dependencies between
properties of different presentable objects, e.g., relative
location Super-classes of Domain-specific properties (binary
relations) Domain-specific classes (n-ary relations) Reflected by
3D content-specific properties 15 statuette.standsOn=stand
cart.isInside=granary
Slide 16
The SEMIC Approach 16 Creation of a generalized content
representation Query-based content generation Building a final
content representation Content presentation Semantic 3D meta-scene
Customized content representation Final content representation 3D
Content Representation Language StepsContent modelsContent
representations Semantic Content Customization Patterns Semantic
Content Model
Slide 17
Semantic 3D Meta-scene Semantic 3D meta-scene a customizable 3D
content representation (knowledge base) generated within the SCM:
Flexible represents 3D content at both low and high levels of
abstraction Generalized represents a super-set of the presentable
content Abstract does not need to specify all elements that are
required for final content presentation Extensible new elements may
be added to a meta-scene to create the final 3D scene 17
Slide 18
Semantic Content Query Semantic Content Query a knowledge base
compliant with the semantic web standards and built according to
Semantic Content Customization Patterns: content selection content
projection content extension content composition 18
Slide 19
Content Customization Patterns Selection Indication of POs,
which are to be included in the generated scene Individual selector
explicit selection of POs Class selector implicit selection of POs
Sub-class selector all POs that belong to a class Restriction
selector selecting POs with a common feature Selector rule
semantically complex queries Complex class selector set operators
(intersection, difference) 19
Slide 20
Content Customization Patterns Selection 20
Slide 21
Content Customization Patterns Projection Indication of
desirable features or behavior of POs that are to be presented in
the generated scene Sub-projector presenting a particular
property/binary RL (or properties/binary RLs with a common root
concept) for all POs included in the generated scene
Super-projector explicit choice of a property/binary RL for
specific POs Conditional projector selection by rules 21
Content Customization Patterns Extension and Composition
Introduction of new content elements that are responsible for new
features or behavior of the selected POs Content extension
introduction of new properties for describing selected POs Content
composition introduction of new RLs for combining selected POs by
specifying mutual dependencies Explicit and implicit content
extension and composition 23
35 Example 2 City models 35 3D meta-scenes Customized 3D scenes
Only palm trees Only privileged cars
Slide 36
36 Example 2 City models 36 3D meta-scenes Customized 3D scenes
Only cars with high beam Only warning and prohibition signs and
passenger cars
Slide 37
Conclusions and Future Works Advantages of the presented
approach Complex content customization mechanisms Operation of a
high-level of abstraction Use of indirect knowledge Platform- and
standard- independence Compliance with the semantic web standards
Future works Implementation of remaining content generation
patterns in the authoring tool Evaluation in terms of size and
complexity of 3D content On-demand generation of rule-based
descriptions of content behavior Support for the four-dimensional
space (including time) Support for context personalization and
adaptation of 3D content 37
Slide 38
Thank you for your attention Semantic Query-based Generation of
Customized 3D Scenes Krzysztof Walczak, Jakub Flotyski Department
of Information Technology Pozna University of Economics, Poland
[walczak, flotynski]@kti.ue.poznan.pl